With the current rate of Intel CPU performance increases, could AMD be catching up?

Page 16 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Homeles

Platinum Member
Dec 9, 2011
2,580
0
0
2. Actually no, but that's a common mis-belief. See this and the following posts after that.
IDC nuked point number one, so let me take care of number two.

AMD's TDP for their 8 core FX series is 125w. That chart you've just linked puts a 4.6 GHz Ivy Bridge at under a 125w power draw. If we keep using the 4-bin boost that their 3770k uses, that brings them up to 4.2GHz with a 4.6GHz turbo, while still being under 125w. With a normal 3770k being clocked at 3.5GHz with a 3.9GHz turbo, that leaves Intel with an 18-20% performance improvement.

Why can you not see this? Intel can get up to 20% more performance with next to zero effort, and they'd still be the performance per watt leader by and large. Yeah, it'd hurt their profits, but there goes your silly claim.
 

Sleepingforest

Platinum Member
Nov 18, 2012
2,375
0
76
I don't think two mobile veterans are quite good enough to counter Intel's large TDP and IPC lead, especially since Intel already has more people working on the mobile space and for longer.

I mean, Clover trail already "manages a huge advantage over NVIDIA's Tegra 3"across the board. Atom is faster than Krait as well, while remain competitive on the TDP side. There's also a Clover Trail+ in the works to counter Tegra 4.
 
Last edited:

Idontcare

Elite Member
Oct 10, 1999
21,110
64
91
I remember us discussing that a while back.
Don't you agree tho - when seeing the Samsung brand as a whole.
It's far more important to (Samsung) damage\sabotage Apple - than having an internal foundry unit thats "kewl"?
They make alot more direct profit\revenue out of this.

Without apple - everything would be samsung in the mobile arena.
Scary stuff.

I think Samsung execs did a costs-benefits analysis and concluded they could have their cake and eat it too if they successfully compete with their customers. Loaded fabs with internally designed ICs and channels full of their vertically integrated products.
 

Fjodor2001

Diamond Member
Feb 6, 2010
4,224
589
126
AMD's TDP for their 8 core FX series is 125w. That chart you've just linked puts a 4.6 GHz Ivy Bridge at under a 125w power draw. If we keep using the 4-bin boost that their 3770k uses, that brings them up to 4.2GHz with a 4.6GHz turbo, while still being under 125w. With a normal 3770k being clocked at 3.5GHz with a 3.9GHz turbo, that leaves Intel with an 18-20% performance improvement.

Why can you not see this? Intel can get up to 20% more performance with next to zero effort, and they'd still be the performance per watt leader by and large. Yeah, it'd hurt their profits, but there goes your silly claim.

The FX series is not the AMD CPU line that correspond to the 3770K. If anything it's A10-5800K @ 100W TDP.

Also, you call Intel potentially going from 77W->125W on a frequency boosted 3770K for a mere 18% performance increase an "easily achievable performance increase"? You really think Intel would even consider going in that direction, given the full focus on lowering power consumption that they currently have?
 

Homeles

Platinum Member
Dec 9, 2011
2,580
0
0
Missed this, AMD will rule on tablet market.
http://hexus.net/business/news/components/50797-amd-recruits-two-high-level-mobile-chip-veterans/

Even if Intel give higher clocks then they will need to spend more on coolers and put a higher price and also there will be less space for overclocking...

Oh, please. Saying AMD will make a comeback an rule the desktop market again is one thing. Saying they're rule the tablet market? Seriously, you are just one big ball of hyperbole. You strike me as one of the guys that writes sensationalist headlines.

Study: "Link Found Between Higher Water Intake and Lower Rates of Cancer"
You: "Drinking More Water Cures Cancer"
The FX series is not the AMD CPU line that correspond to the 3770K. If anything it's A10-5800K @ 100W TDP.
The performance gap there is even larger. That's not working out too well for your side of the fight, is it?
Also, you call Intel potentially going from 77W->125W on a frequency boosted 3770K for a mere 18% performance increase an "easily achievable performance increase"? You really think Intel would even consider going in that direction, given the full focus on lowering power consumption that they currently have?
You think Intel's going to let AMD just waltz in and take the performance crown? If they have to release a faster part, they will.
 
Last edited:

vampirr

Member
Mar 7, 2013
132
0
0
Oh, please. Saying AMD will make a comeback an rule the desktop market again is one thing. Saying they're rule the tablet market? Seriously, you are just one big ball of hyperbole.

AMD has superior technology for the tablet market thanks to APU's that can deliver a great bang for a buck and since they have now three veterans that can aid in development for the tablet market we can expect great products for the price, performance, power consumption and TDP in one package...

AMD will enter ARM in 2014 and since they have needed technology, design it can aid their ARM architecture with APU inspired design and Jim Keller can help in ARM research and development of AMD's ARM based products.

They have three guys that can have some noticable effect on AMD design's, research and products...
 

Sleepingforest

Platinum Member
Nov 18, 2012
2,375
0
76
AMD has superior technology for the tablet market thanks to APU's that can deliver a great bang for a buck and since they have now three veterans that can aid in development for the tablet market we can expect great products for the price, performance, power consumption and TDP in one package...

AMD will enter ARM in 2014 and since they have needed technology, design it can aid their ARM architecture with APU inspired design and Jim Keller can help in ARM research and development of AMD's ARM based products.

They have three guys that can have some noticable effect on AMD design's, research and products...

Okay, do you know what already has that? Terga, Apple's A#, Krait, and now Clover Trail+. Yes, those guys will be helpful. But AMD is entering the game late and with less personnel specializing in low TDP than any other SOC developer.

AMD stands a chance, but it's even less than Intel simply because it's budget is already stretched thin and they're generally leaking talent.
 

pm

Elite Member Mobile Devices
Jan 25, 2000
7,419
22
81
I think Samsung execs did a costs-benefits analysis and concluded they could have their cake and eat it too if they successfully compete with their customers. Loaded fabs with internally designed ICs and channels full of their vertically integrated products.

I agree with that, but I'd point out that that dance is not quite as easy as you make it sound. It's great to have fabs full of products that you can use and you can sell to your customers but that only works until you run out of supply and have to decide between your own products and your customers. If you treat your customers ok, you lose margin, but if you tell your customers that they have no wafer starts, then they walk. I have to give Samsung credit for mostly making it work - and making it seem easy.
 

SomeoneSimple

Member
Aug 15, 2012
63
0
0
AMD has superior technology for the tablet market thanks to APU's that can deliver a great bang for a buck [..]

I know, right? Just like their Notebook APU platform, ... where AMD has an even lower market penetration compared to their desktop solutions.
 

Exophase

Diamond Member
Apr 19, 2012
4,439
9
81
The intrinsic cost per wafer for TSMC is rising, but no more so than has been the traditional node-on-node increase.

TSMC basically has no competition. It is up to them to use free-market management methods to determine what prices the market will bear.

If GloFo or UMC were anywhere close to being competitive (in reality I mean, not just in PR land with powerpoint files) then TSMC's prices would have to deflate to reflect that customers would haggle a lot more aggressively.

Thank you for the explanation, I agree this makes sense. It's sad that GF can't keep up, and are only now finally releasing 28nm mobile products. That'd put them about a year behind TSMC, which is substantial. I just hope that increased margins and increased volume, particularly from Apple, can help them stay closer to Intel than they might have otherwise (even if still falling even further behind)

Samsung was the closest thing to competition for TSMC but their legal dept went and shot their foundry division in the foot by very publicly suing a foundry customer (Apple) in addition to simultaneously having their internal handset development team compete head-to-head with their foundry customer (Apple). You don't do that if you want your existing customers to stick around, or potential customers to come onboard.

Does Samsung even have any other high profile clients for its leading edge processes outside of Apple and their SoC division? Is there any cross benefit at all between their DRAM and NAND processes and mainline logic processes? I know things are usually very different but don't know if there's zero common in R&D and equipment or not.. I figure you can answer this though.

To be fair, it was Apple that sued Samsung, not the other way around, and Samsung doing phones is nothing new. Maybe you think that they should have settled? But Apple is really fickle, I don't even know if the lawsuit is their full reason for wanting to leaving. I don't think it really matters for other customers either, because no potential SoC manufacturer is nearly as vertically integrated as Apple or Samsung themselves are. In fact, if Qualcomm or nVidia moves to Samsung (and both have expressed concern with TSMC) that could work to everyone's favor since Samsung has used both in their phones.
 

2is

Diamond Member
Apr 8, 2012
4,281
131
106
I know, right? Just like their Notebook APU platform, ... where AMD has an even lower market penetration compared to their desktop solutions.

Just need to ignore him. Nobody actually takes him seriously. His most recent "amd rules" thread was largely ignored and turned into a chicken song (youtube) benchmarking contest to see who can have the most tabs open before the browser or plugin crashed.
 

AtenRa

Lifer
Feb 2, 2009
14,003
3,362
136
IDC nuked point number one, so let me take care of number two.

AMD's TDP for their 8 core FX series is 125w. That chart you've just linked puts a 4.6 GHz Ivy Bridge at under a 125w power draw. If we keep using the 4-bin boost that their 3770k uses, that brings them up to 4.2GHz with a 4.6GHz turbo, while still being under 125w. With a normal 3770k being clocked at 3.5GHz with a 3.9GHz turbo, that leaves Intel with an 18-20% performance improvement.

Why can you not see this? Intel can get up to 20% more performance with next to zero effort, and they'd still be the performance per watt leader by and large. Yeah, it'd hurt their profits, but there goes your silly claim.

In order for the 3770K to reach the 4.2GHz and 4.6GHz Turbo (single core) Intel will need to sell it with a BIG Heat-Sink Fan(or water cooling) in order to keep it at 105c Tjmax. That means higher product cost and thus lower gross margins.

Also as IDC have said before, they will cannibalize its own Socket 2011 products by doing this.
 

Dark Shroud

Golden Member
Mar 26, 2010
1,576
1
0
I know, right? Just like their Notebook APU platform, ... where AMD has an even lower market penetration compared to their desktop solutions.

Yes AMD has lower numbers, that doesn't change the fact that their Bobcat chips were better than Atom and now their Jaguar chips are even better than Bobcat.

For some reason OEMs kept cramming Bobcat chips into larger laptops than they should have. I've used Bobcat Netbooks and they're really good, only problem is almost no one made them. Yet plenty of people touted lesser power Atom netbooks.

I'm not holding my breath for an AMD ARM SOC. But I do expect it to perform well.
 

Fjodor2001

Diamond Member
Feb 6, 2010
4,224
589
126
The performance gap there is even larger. That's not working out too well for your side of the fight, is it?

You think Intel's going to let AMD just waltz in and take the performance crown? If they have to release a faster part, they will.

The discussion was whether Intel "easily could adjust the frequency upwards on existing CPUs to gain a substantial performance increase with few drawbacks".

If it would mean Intel would go from 77W->125W on a 3770K for a mere 18% performance increase then I guess it's pretty obvious that it's not a reasonable way for Intel to easily increase their CPU performance. I.e. they do not have much headroom to increase the frequency without taking a serious TDP penalty.

It's actually pretty obvious that Intel has hit a frequency wall by looking at the latest CPU releases. SB (2600K) -> IB (3770K) has only brought 100 MHz frequency increase over the 2.5 years that passed between those two CPU releases. It's because Intel knows that increasing the frequency of their current CPUs means a too high TDP penalty to be worth considering (and also because they focus on lowering TDP).
EDIT: "IB (3770K)" should be "Haswell (4770K)".

If it really would be easy for Intel to increase the CPU frequency with few drawbacks, don't you think they already would have done so?
 
Last edited:

mrmt

Diamond Member
Aug 18, 2012
3,974
0
76
If it would mean Intel would go from 77W->125W on a 3770K for a mere 18% performance increase then I guess it's pretty obvious that it's not a reasonable way for Intel to easily increase their CPU performance. I.e. they do not have much headroom to increase the frequency without taking a serious TDP penalty.

Don't forget that to beat the FX8350 they could go all the way up to 140W, not 125W. It's more like 25% headroom, not 18%.
 

SomeoneSimple

Member
Aug 15, 2012
63
0
0
If it would mean Intel would go from 77W->125W on a 3770K for a mere 18% performance increase then I guess it's pretty obvious that it's not a reasonable way for Intel to easily increase their CPU performance. I.e. they do not have much headroom to increase the frequency without taking a serious TDP penalty.

power-3.png


Yeah, you tell em. Intel has no headroom for higher clocks, and no TDP headroom either, at all. AMD really has the edge here. :awe:
 
Last edited:

Fjodor2001

Diamond Member
Feb 6, 2010
4,224
589
126
The intrinsic cost per wafer for TSMC is rising, but no more so than has been the traditional node-on-node increase.

TSMC basically has no competition. It is up to them to use free-market management methods to determine what prices the market will bear.

If GloFo or UMC were anywhere close to being competitive (in reality I mean, not just in PR land with powerpoint files) then TSMC's prices would have to deflate to reflect that customers would haggle a lot more aggressively.

Samsung was the closest thing to competition for TSMC but their legal dept went and shot their foundry division in the foot by very publicly suing a foundry customer (Apple) in addition to simultaneously having their internal handset development team compete head-to-head with their foundry customer (Apple). You don't do that if you want your existing customers to stick around, or potential customers to come onboard.

You may have a point that TSCM having little competition may lead to increased price for its customers.

But I guess that's not directly a node advantage as such for Intel, more of an advantage due to lack of competition. If the end result means increased costs for Intel's competitors, then I agree that it still is a valid argument for being a benefit for Intel though.

However, it is uncertain just how much TSCM can increase the price due to lack of competition. 10%, 15%, 30%? Hard to know. Then we also have for example the possibility that GloFo might get it's 28 nm process tech up and running for use in Kaveri which may bring prices down. And that Intel by itself will have to bear ever increasing R&D cost for developing new node tech. In other words there are lots of factors involved in determining what the price actually will be, who bears what R&D costs, who gets what volumes, etc. So the sum of this and the magnitude of its effect is hard to estimate.

Then we also have the following important aspects:

-Going from 32->22 nm did not add much room for increased CPU frequency. It is reasonable to assume that this will also be true when going from 22->14nm.

-Going to a later node will likely provide lower TDP. But SB(2600K)->IB(3770K) "only" meant 95W-77W=18W TDP reduction (to be fair, the iGPU also grew a bit adding to the transistor count). So the difference is not huge.

Based on all of this I would say that the advantage that Intel has of being on a later node is getting less after all, compared to how it has been in the past.
 

AtenRa

Lifer
Feb 2, 2009
14,003
3,362
136
Don't forget that to beat the FX8350 they could go all the way up to 140W, not 125W. It's more like 25% headroom, not 18%.

At 140W TDP they will need a WaterCooling in order to keep the TJmax at 105C.

Just to give you an example,

Core i7 3770K at 3.8GHz(All cores) with the default Heat-Sink fan at 35C ambient temperature will hit the 105C TJmax and throttle down. In order to raise the frequency Intel will need to sell the CPU with more expensive CPU Coolers. They will never do that.
 

Fjodor2001

Diamond Member
Feb 6, 2010
4,224
589
126
Don't forget that to beat the FX8350 they could go all the way up to 140W, not 125W. It's more like 25% headroom, not 18%.

The FX8350 has 125 W TDP:

http://en.wikipedia.org/wiki/List_of_AMD_FX_microprocessors#.22Vishera.22_.2832_nm_SOI.29

Also, the FX8350 is not the AMD CPU that most resembles the 3570/3770K. The FX-8350 is not even an APU like the 3570/3770K. If anything you should be comparing to the A10-5800K @ 100W TDP. But there really is no exact corresponding AMD CPU that compares directly to the 3570K/3770K due to their differences uarch design (AMD's use of Modules and such).

But anyway, you're still missing the main point, which is that Intel CPUs have very little headroom to increase the CPU frequency without taking a massive TDP penalty hit. 77W->125W TDP increase for a mere 18% performance improvement is insane, and something neither Intel nor AMD would ever opt for.
 
Last edited:

SomeoneSimple

Member
Aug 15, 2012
63
0
0
The FX8350 has 125 W TDP:

125W TDP on the FX8350 is bollocks.

http://www.heise.de/newsticker/meldung/AMD-Acht-Kerne-und-4-GHz-kontra-Intels-Core-i-1734298.html

Translation
Power: Our test processor swallowed up to 168 watts - we measured the 12-volt supply line between the power supply and motherboard. Even with the loss of the VRM's on the board calculated generously, it is clear that the 125 watt TDP is pure wishful thinking.

I also absolutely agree, that the A10-5800K is a MUCH better comparison to the 3770K, than the 8350. :awe:

50409.png